This paper consists a data reconstruction method in an Unmanned Aerial Vehicles (UAV)'s propeller speed measurement Revolution Per Minute (RPM) that is contaminated by outliers. This method will be suitable for UAV modeling, system identification or controlling purposes which required reliable data of propeller RPM from its data acquisition system. The method predicts the probable RPM value from its valid past data and use that prediction to detects the outliers out of the available measurement. Previously, the existing outlier's detection methods were found out to be too sensitive and detect too much false outliers of the following process although these were valid data. On the other hand, the decrement of sensitivity causing failures in detection process and yielding too little outliers. Hence, this UAV propeller's RPM filtering method modified the prediction of probable value of the future RPM and smoothing the filtering process and detect all the outliers out from RPM data stream.